Cokriging random fields with means related by known linear combinations

Traditional approaches to predict a second-order stationary vector random field include simple and ordinary cokriging, depending on whether or not the mean values of the vector components are assumed to be known. This paper explores a variant of cokriging, in which the mean values of the vector components are related by linear combinations with known coefficients. Equations for the cokriging predictor and for the variance–covariance matrix of prediction errors are presented. A set of computer programs is provided and illustrated with applications to mineral resources evaluation, in which the proposed cokriging variant compares favorably with traditional approaches.

► Cokriging relies on the modeling of mean values and spatial correlation structure. ► Improvements are obtained by accounting for linear dependence between mean values. ► Computer programs are provided and illustrated with mining data sets.